【作者】 沈天飞；

【导师】 陈伯时； 龚幼民；

【作者基本信息】 上海大学 ， 控制理论与控制工程， 2001， 博士

【摘要】 我国目前大部分起重机的拖动电机是绕线式异步电动机，一般采用有级开环调速方法，为了适应现代化生产的要求，亟待进行技术改造，在有些应用场合，转子斩波调阻调速法是一种比较经济实用的方法。基于可控硅的绕线式电机的转子斩波调阻法是一个已经成熟的技术，为了适应电力电子技术的发展，本文探讨了基于IGBT的转子斩波调阻调速法。采用IGBT的斩波电路可以省去原来可控硅电路所需的平波电抗器和强迫关断电路，因此斩波电路结构可以得到极大的简化。文中提出了以IGBT、二极管、电阻和电解电容等四个元器件组成的斩波电路，经实验和仿真证明不仅能有效地调节转子电阻，而且同时兼有IGBT关断吸收电路的作用，能显著改善IGBT的关断轨迹。本文建立了该调速系统的准动态数学模型，首次推导出了IGBT的占空比与斩波电路等效电阻之间的非线性关系，并给出了该斩波回路的工程设计方法，确保了所选择的电阻和电容值不仅能使系统达到指定的调速范围，而且还能使IGBT在关断时避免承受功率尖峰。鉴于有些应用场合需要有轻载低速的工作点，本文在转子斩波调阻的基础上，增加了定子降压调速，推导出了定子电压的计算公式。本文还给出了该系统的动态结构框图，提出了电流环和转速环的PI调节器的设计方法，实现了起重机的无级调速和闭环控制。 为了用电路仿真软件PSPICE辅助本课题的研究工作，本文以IGBT为例研究了PSPICE中组合模型特性参数的辨识问2(X)1年上海大学博士学位论文题，并成功地将BP神经网络应用于IGBT组合模型静态特性参数的辨识，使得器件手册上的静态特性可以通过人工神经网络的运算方便地转换为相应的组合模型参数，为组合模型参数的辨识寻找到了一条捷径。 鉴于用于IGBT组合模型静态特性参数辨识的BP网络在训练时存在很多局部极小点的情况，本文提出了一种改进的自适应变步长梯度算法(IAVSG)，使网络避免了陷人局部极小点。采用该算法后，网络训练时的收敛速度大为提高。更多还原

【Abstract】 Wound-rotor induction motors are used in a large amount of cranes in our country, and are controlled in discrete steps and open loop. These equipments are not suitable in some areas nowadays. So their reforming has become quite urgent. In some cases, rotor resistance chopper control scheme is economic and efficient. The thyristor- chopper control scheme has already been investigated thoroughly. This paper probes into the rotor resistance IGBT chopper control system for the induction motor to follow the development of modern power electronics. The IGBT-controlled chopper circuit can be greatly simplified as the smoothing inductor and commutation circuit of the thyristor-controlled chopper circuit are no more required. In this paper a novel chopper circuit consisting only four elements: one IGBT, one diode, one resistor and one electrolytic capacitor is presented. Experiments and simulation have shown that the circuit can not only adjust the resistance in the rotor circuit efficiently, but also absorb the spikes across IGBT when the switch is turned off, and the turn-off track of IGBT can be improved a lot. The equivalent circuit of the motor with this circuit has been analyzed in detail, and the nonlinear mapping between the equivalent resistance and the duty cycle is deduced. The value of the resistor and capacitor can be determined according to the design method given in the paper so that the specified speed ratio can beachieved and the power spikes across IGBT when it is turned-off can be weakened or avoided. In our system, the stator voltage control is also applied as a secondary strategy, so that the motor can also run at low speeds with light load. The practical calculation method of the stator voltage has also been introduced.In order to simulate the above system with old version of PSPICE, we have investigated the identification of model parameters of IGBT composite model in this software, and BP neural network has been applied. With the help of neural network, the steady-state behavior related parameters of IGBT can be acquired quite easily.During the training of the BP network in this application, we have observed that there exist a large amount of local minimal points. We have developed an unproved adaptive variable step gradient (IAVSG) algorithm, so that the net can jump out of the local minimums more efficiently, and the convergence speed can be greatly increased.更多还原